U.S. patent application number 14/126142 was filed with the patent office on 2016-05-12 for type error revising method.
The applicant listed for this patent is KEUKEY INC.. Invention is credited to Min Chul KIM.
Application Number | 20160132129 14/126142 |
Document ID | / |
Family ID | 49220119 |
Filed Date | 2016-05-12 |
United States Patent
Application |
20160132129 |
Kind Code |
A1 |
KIM; Min Chul |
May 12, 2016 |
TYPE ERROR REVISING METHOD
Abstract
The present invention is directed to a type error revising
method. According to the present invention, a new and innovative
error revising method is provided, which does not require
inconvenient action of inputting deletion function key like
backspace or moving a cursor by the mouse control or finger touch.
The type error revising method of the present invention comprises
steps of; a user inputs revision target word or character substring
in the current cursor location; the user's revision intention is
determined automatically or according to user's input of revision
command; and one or more words or character substrings input before
the revision target word or character substring are replaced by the
revision target word or character substring. In the present
invention, the replaced one or more words or character substrings
are selected through a correlation calculation with the revision
target word or character substring or substring thereof.
Inventors: |
KIM; Min Chul; (Seoul,
KR) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
KEUKEY INC. |
Seoul |
|
KR |
|
|
Family ID: |
49220119 |
Appl. No.: |
14/126142 |
Filed: |
November 21, 2013 |
PCT Filed: |
November 21, 2013 |
PCT NO: |
PCT/KR2013/010595 |
371 Date: |
December 13, 2013 |
Current U.S.
Class: |
345/168 |
Current CPC
Class: |
G06F 3/0237 20130101;
G06F 3/033 20130101; G06F 40/242 20200101; G06F 3/0233 20130101;
G06F 40/232 20200101 |
International
Class: |
G06F 3/023 20060101
G06F003/023; G06F 3/033 20060101 G06F003/033; G06F 17/27 20060101
G06F017/27 |
Foreign Application Data
Date |
Code |
Application Number |
Apr 15, 2013 |
KR |
10-2013-0041134 |
Claims
1. A correction method of word or character substring comprising
the steps of; a user inputs revision target word or character
substring in the current cursor location; the user's revision
intention is determined automatically or according to user's input
of revision command; one or more words or character substrings
input before the revision target word or character substring are
replaced by the revision target word or character substring;
wherein the replaced one or more words or character substrings are
selected through a correlation calculation with the revision target
word or character substring or substring thereof.
2. A correction method of word or character substring according to
claim 1, wherein the user's revision intention is automatically
determined in the step of the user's revision intention
determination step, and the automatic user intention determination
step comprises steps of; a comparison with the library is carried
out every time when the user inputs word; word which is not listed
in the library is registered as revision candidate; a correlation
between the revision target word or character substring and the
revision candidate is calculated; when the calculated correlation
is larger than a predetermined value, the user's input of revision
target word or character substring is determined to be made out of
revision intention.
3. A correction method of word or character substring according to
claim 1, wherein the user's revision intention is determined
according to the user's input of revision command, and the subject
of the correlation calculation with the revision target word or
character substring is one or more words located before the
revision target word or character substring.
4. A correction method of word or character substring according to
claim 1, wherein the user's revision intention is automatically
determined in the step of the user's revision intention
determination step, and the automatic user intention determination
step comprises steps of; a comparison with the library is carried
out every time when the user inputs word; word which is not listed
in the library is registered as a revision candidate; a correlation
between a substring starting from the first character of the
revision target word or character substring and the revision
candidate, which is input right before the revision target word or
character substring, is calculated; wherein the word, which is
located right before the revision target word or character
substring, is replaced in the replacement step.
5. A correction method of word or character substring according to
claim 1, wherein the user's revision intention is determined
according to the user's input of revision command, the word, which
is located right before the revision target word or character
substring, is replaced in the replacement step, and the correlation
calculation is carried out between a substring starting from the
first character of the revision target word or character substring
and the word, which is located right before the revision target
word or character substring.
Description
TECHNICAL FIELD
[0001] The present invention is directed to a type error revising
method and a device implementing the same method. The method of the
present invention can be applied to a keyboard input of personal
computers as well as to a touch keypad input of smart phones and
keypad input of all mobile phones (including feature phones).
BACKGROUND ART
[0002] Typical type error revising method when inputting characters
(all types of inputtable characters including English characters,
Korean characters, numbers, special characters, etc.) in personal
computers or in mobile phones is to delete all words between the
word including type error and the current cursor location using
deletion function key like backspace, and retype the whole deleted
words, or to move a cursor to the word including type error using a
mouse (in case of personal computers) or using a finger touch (in
case of mobile devices) and retype the word including type
error.
[0003] However, when there exist correctly typed words between the
word including type error and the current cursor location, the
correctly typed words as well as the word including type error
should be deleted when applying the typical type error revising
method of inputting deletion function key like backspace. When
moving a cursor using a mouse in personal computers, one of the
hands should be relocated from a keyboard to a mouse, the cursor
should be accurately placed in a location of type error through the
mouse control, and the cursor should be returned to the previous
location for resuming character input through the mouse control.
This user attention requiring process is quite inconvenient. When
moving a cursor using a finger touch in mobile devices, the
accuracy of cursor control is not generally good enough to locate
the cursor in a location of type error at one time and also
requires user attention, thus is quite inconvenient.
[0004] The auto-correction function was developed in order to
enhance user convenience in revising type errors. This
auto-correction function has been already commercialized, and works
to automatically correct the word "A," which the user actually
typed, to the word "B," when the machine decides that the word "A"
is a type error and should be revised to the word "B."
Automatically retyped word "B" should exist in a library of the
machine, and is not required to be retyped by the user.
[0005] In the above auto-correction function, it is apparent that
the retyped word "B" is not actually input by the user with the
intention of revision, but is what the machine presumes that the
user actually intended to input. Thus, there are problems when the
accuracy of presumption is not good enough. The problematic
accuracy of presumption is partly because of the limited number of
words registered in a library of the machine. These days, many
users frequently use newly coined words or new abbreviations, which
are not covered by the old library, in typing environment of
personal computers and mobile devices. Thus, if the library is not
frequently updated, the limitation of the library becomes more and
more remarkable. For this reason, many users turn the
auto-correction function off when they input characters, even
though their devices (personal computers or mobile devices) are
equipped with the auto-correction function.
[0006] A further problem of the auto-correction function is that
the auto-correction works only in a currently inputting word. Thus,
when the users find that there are type errors in the already input
words, not in the currently inputting words, like in the typical
error revising methods described above, the words between the word
including type errors and the current cursor location should be
deleted through inputting the deletion function key like backspace,
or a cursor should be moved by a mouse control or finger touch to
the location of type errors.
DISCLOSURE OF INVENTION
Technical Problem
[0007] The objective of the present invention is to solve the
problems of the above-described prior art technologies. More
specifically, the objective of the present invention is to provide
an error revising method and a device implementing this method,
which does not require inconvenient action of inputting deletion
function key like backspace or moving a cursor by the mouse control
or finger touch.
Solution to Problem
[0008] The unique technical means, which achieve the objectives of
the present invention, will be described in the following.
[0009] The technical means of the present invention can be
classified into three (3) groups, and each group can be classified
into a method in which user's intention for revision is
automatically detected and a method in which user directly provides
a revision command.
[0010] It is common in all means of the present invention that the
type errors are not deleted using the deletion function key like
backspace, and that the cursor is not controlled by the user to be
located to an accurate position of the type errors. The user does
not move the cursor location for the type error revision, but
inputs correct words or characters without cursor location change.
The user's activity for revising type errors takes place in the
current cursor location.
[0011] In the specification of the present invention, the term
"word" is defined as series of characters. It is not required that
the word, i.e., series of characters is listed in the library. Each
word is separated from another word by a space.
[0012] (First means: Comparison between the one or more already
input word and the newly input word in a predetermined range)
[0013] This means is applied when there are type errors in one or
more already input words. One or more words including type errors
are replaced with newly input word.
[0014] First, the method, in which user's intention for revision is
automatically detected, is described.
[0015] Every time when the user inputs word, the input word is
compared with a library. The word which is not listed in the
library is registered as a candidate word to be revised. A
correlation between the newly input word and the candidate word to
be revised is calculated. If the correlation is larger than
predetermined value, then it is decided that the newly input word
is made out of the user's intention for revision. When a space is
input after the newly input word, then the candidate word to be
revised is replaced by the newly input word.
[0016] Next, the method, in which user directly provides a revision
command, is described.
[0017] If the user inputs word and thereafter a revision command is
made, the input word is recognized as a revision target word. Then,
a correlation between the revision target word and the words, which
have been input before the input of the revision target word, is
calculated. If there exist one or more words, of which the
calculated correlation is above predetermined value, such words are
replaced by the revision target word.
[0018] (Second means: Comparison between the newly input word and
the word input right before the newly input word)
[0019] This means is applied when the user recognizes type error
before completing input of the intended word. In this case, the
user inputs a space in a state that input of the intended word is
not completed, inputs a new correct word, and then replaces the
word including type error input right before the new correct word
with the new correct word.
[0020] First, the method in which user's intention for revision is
automatically detected is described.
[0021] Every time when the user inputs word, the input word is
compared with a library. The word which is not listed in the
library is registered as a candidate word to be revised. A
correlation between a substring starting from the first character
of the newly input word and the candidate word, which is input
right before the newly input word, is calculated. If the
correlation is larger than predetermined value, then it is decided
that the newly input word is made out of the user's intention for
revision. When a space is input after the newly input word, then
the candidate word to be revised is replaced by the newly input
word.
[0022] Next, the method in which the user directly provides a
revision command is described.
[0023] If the user inputs word and thereafter a revision command is
made, the input word is recognized as a revision target word. Then,
a correlation between a substring starting from the first character
of the revision target word and the word, which is input right
before the revision target word, is calculated. If the correlation
is larger than predetermined value, then the word, which is input
right before the revision target word, is replaced by the revision
target word.
[0024] (Third means: Comparison between substrings within a
word)
[0025] This means is applied when the user recognizes type error
before completing input of the intended word. Differently from the
second means, the user does not input a space and continuously
input correct characters and then the character substring including
type error is replaced by the correct character substring to
complete input of a word.
[0026] First, the method, in which user's intention for revision is
automatically detected, is described.
[0027] If the user completes input of a word, the input word is
compared with a library. In case the input word is not listed in
the library, it is determined whether there exist character
substrings within the input word, which are highly correlated with
each other. If it is determined that there exist character
substrings within the input word, which are highly correlated with
each other, then it is determined whether a word, in which the
previously input character substring is replaced by the later input
character substring, exists in the library. If it is determined
that such word exists in the library, the previously input
character substring is replaced by the later input character
substring to complete input of a word.
[0028] Next, the method, in which user directly provides a revision
command, is described.
[0029] If the user inputs word and thereafter a revision command is
made, a correlation between character substrings within the input
word is calculated. If it is determined that there exist character
substrings within the input word, which are highly correlated with
each other, then it is determined whether a word, in which the
previously input character substring is replaced by the later input
character substring, exists in the library. If it is determined
that such word exists in the library, the previously input
character substring is replaced by the later input character
substring to complete input of a word.
Advantageous Effects of Invention
[0030] According to the present invention, it becomes possible to
correct type errors without inputting deletion function keys such
as a backspace, thus a fast and convenient error correction becomes
possible in comparison with the conventional method of type error
correction.
BRIEF DESCRIPTION OF DRAWINGS
[0031] FIG. 1 depicts an embodiment in which user's intention for
revision is automatically detected according to the first
means.
[0032] FIG. 2 depicts an embodiment in which user directly provides
a revision command according to the first means.
[0033] FIG. 3 depicts an embodiment in which user's intention for
revision is automatically detected according to the second
means.
[0034] FIG. 4 depicts an embodiment in which user directly provides
a revision command according to the second means.
[0035] FIG. 5 depicts an embodiment in which user's intention for
revision is automatically detected according to the third
means.
[0036] FIG. 6 depicts an embodiment in which user directly provides
a revision command according to the third means.
MODE OF THE INVENTION
[0037] FIG. 1 depicts an embodiment in which user's intention for
revision is automatically detected according to the first means.
The user intended input of "Happy Easter everyone," but actually
input "Hsppy Easter wveryone." According to the first means, every
time when the user inputs word, the input word is compared with a
library, and the word which is not listed in the library is
registered as a candidate word to be revised. Thus, in the above
case, the words "Hsppy" and "wveryone" are registered as candidate
words to be revised. At the state that the user has input "Hsppy
Easter wveryone," the user recognizes type error made at
"wveryone," and wishes to revise it to "everyone." For this, the
user inputs a space after "Hsppy Easter wveryone," and then inputs
"everyone." At this state, the user's input is now "Hsppy Easter
wveryone everyone." According to the first means, a correlation
between the newly input word and the candidate word to be revised
is calculated. In this example, since the newly input word,
"everyone," is highly correlated with "wveryone," which is one of
the candidate words to be revised, the newly input word,
"everyone," is determined to have been made out of the user's
intention for revision. According to such determination, when the
users inputs a space, "wveryone" is replaced by "everyone" so that
the whole sentence "Hsppy Easter wveryone everyone" is changed to
"Hsppy Easter everyone." At this state, the user additionally
recognizes type error made at "Hsppy," and wishes to revise it. For
this, the user inputs a correct word of "happy." Since one of the
two candidate words to be revised, "wveryone," has been already
revised to "everyone," there currently exists only one candidate
word to be revised, which is "Hsppy." Since "Hsppy" is highly
correlated with "happy," the newly input word, "happy," is
determined to have been made out of the user's intention for
revision. According to such determination, "Hsppy," which is a
candidate word to be revised, is replaced by "happy," and in the
course of such replacement, it is further detected that the word
"happy" should be the first word of a sentence, accordingly a
character "h" of "happy" is changed to an upper case "H." Now, the
whole sentence "Hsppy Easter everyone" is changed to "Happy Easter
everyone" to complete revision. The range of determination of the
candidate words to be revised can be freely set according to the
circumstances. For example, the range may be set to ten (10) words
before the currently inputting word or to hundred (100) words
before the currently inputting word. The range can also be set to
the entire document which the user is working on. The method of
determination of the correlation between words, which is used
during the above revision procedure, is described below.
[0038] A determination of the correlation between words can be
practiced in various ways. Representatively, the following two ways
are used. The first one is based on that although a user intends to
input a certain key, the possibility of actual input of a key
abutting the intended key in a keyboard layout is high. For
example, in a correlation determination between "everyone" and
"wveryone," there exists a difference between the two words only in
the first characters "e," and "w." Also, the character "e" and "w"
abut each other in a keyboard layout, thus it is highly likely that
the user intended to input "e," but the actual input of "w" was
made. Accordingly, the system determines that "everyone" and
"wveryone" are highly correlated with each other. The second one is
based on the changed input sequence of characters. When a user
inputs characters in a high speed, an input sequence of characters
are frequently changed. For example, in a correlation determination
between "everyone" and "evreyone," there exists a difference
between the two words in the third and fourth characters, i.e.,
"er" in "everyone" and "re" in "evreyone." Between "er" and "re," a
difference lies in the input sequence. In such a case, the system
determines that "everyone" and "evreyone," are highly correlated
with each other. When a user inputs a long word, the number of
mis-inputs of abutting keys and the number of mis-inputs of changed
input sequence may increase, thus in a quantification processing of
the correlation, a different standard may be applied according to
the length of a word. The correlation determination methods, which
can be applied in the present invention, are not limited to the
above-described two methods.
[0039] FIG. 2 depicts an embodiment in which the user directly
provides a revision command according to the first means. The user
intended input of "Happy Easter everyone," but actually input
"Hsppy Easter wveryone." The user first recognizes type error at
"wveryone," and for a revision thereof, inputs a space after "Hsppy
Easter wveryone," and then types "everyone." At this state, the
user makes a revision command. The revision command may be
practiced in various ways. Using a function key or a certain area
in a virtual keyboard for the revision command is possible. When
the revision command is practiced in a mobile device, a slight
shaking action of the mobile device, or left, right, upword or
downword swiping gestures in virtual keyboard may be recognized as
the revision command. In any way, if the revision command is made
after the input of "everyone," the system recognizes "everyone" as
a revision target word. Then, a correlation between the revision
target word and the words, which have been input before input of
the revision target word, is calculated. Among the words, which
have been input before the input of "everyone," which is the
revision target word, "wveryone" is highly related with the
revision target word, "everyone." Thus, "wveryone," is replaced by
the revision target word, "everyone," according to the revision
command. At this state, "Hsppy Easter wveryone everyone," which the
user has input until now is changed to "Hsppy Easter everyone." The
user now additionally recognizes a type error at "Hsppy," and
inputs a space after "everyone," and continuously inputs "happy."
Thereafter, the user makes the revision command again. Then,
"happy" is recognized as the revision target word, and a
correlation between the revision target word and the words, which
have been input before input of the revision target word, is
calculated. According to such correlation calculation, "Hsppy" is
found to be highly related to "happy." Thus, "Hsppy" is replaced by
"happy," and in the course of such replacement, it is further
detected that the word "happy" should be the first word of a
sentence, accordingly a character "h" of "happy" is changed to an
upper case "H." Now, the whole sentence "Hsppy Easter everyone" is
changed to "Happy Easter everyone" to complete revision. The range
of determination of the candidate words to be revised can be freely
set according to the circumstances. For example, the range may be
set to ten (10) words before the currently inputting word or to
hundred (100) words before the currently inputting word. The range
can also be set to the entire document which the user is working
on. For the correlation calculation methods, the same methods as
described in the above embodiment of FIG. 1 can be used.
[0040] FIG. 3 depicts an embodiment in which user's intention for
revision is automatically detected according to the second means.
While the user is typing with the intention of typing "I remember,"
the user recognizes that the actual typing has been made as "I
teme" by mistake instead of "I reme". At this state, the user
inputs a space, then "teme" is recognized by the system as a word.
According to the second means, every time when the user inputs
word, the input word is compared with a library, and the word which
is not listed in the library is registered as a candidate word to
be revised. Since "teme" is not in the library, it is registered in
the system as a candidate word to be revised. Thereafter, the user
inputs originally intended word, "remember," thus the user's entire
input is now "I teme remember." A correlation between a substring
starting from the first character of the newly input word, i.e.,
"r" and the candidate word, which is input right before the newly
input word, is calculated. Since the candidate word, which is input
right before the newly input word, "teme," comprises four (4)
characters, the first four (4) characters of "remember," i.e.,
"reme" may be compared with "teme." But the length of substring
staring from the first character of the newly input word is not
limited to the number of characters of the candidate word. Since
the candidate word, "teme," which is input right before the newly
input word, and "reme," which is substring staring from the first
character of the newly input word, are highly correlated with each
other, the system recognizes that the newly input word "remember"
is out of the user's intention for revision. Then, "teme," which is
a candidate word to be revised, is replaced by "remember" to
complete revision. For the correlation calculation methods, the
same methods as described in the above embodiment of FIG. 1 can be
used.
[0041] FIG. 4 depicts an embodiment in which the user directly
provides a revision command according to the second means. While
the user is typing with the intention of typing "I remember," the
user recognizes that the actual typing has been made as "I teme" by
mistake instead of "I reme." At this state, the user inputs a
space, and then continuously inputs "remember." The revision
command is made thereafter. According to the revision command,
which can be practiced variously as described above, "remember" is
recognized by the system as a revision target word. A correlation
between a substring starting from the first character of the
revision target word and the word, which is input right before the
revision target word, is calculated. Since the word, which is input
right before the revision target word, "teme," comprises four (4)
characters, the first four (4) characters of "remember," i.e.,
"reme" may be compared with "teme." But the length of substring
staring from the first character of the revision target word is not
limited to the number of characters of the word, which is input
right before the revision target word. Since "reme" and "teme" are
highly related with each other, "teme," which is input right before
the revision target word, is replaced by "remember," which is
revision target word to complete revision. For the correlation
calculation methods, the same methods as described in the above
embodiment of FIG. 1 can be used.
[0042] FIG. 5 depicts an embodiment in which user's intention for
revision is automatically detected according to the third means.
The user intended to input "I remember," but at the state that the
actual input is made as "I rene," the user recognizes that the
actual input of "rene" was erroneously made instead of correct
input of "reme." In the embodiments of FIGS. 3 and 4, the user
inputs a space to make "rene" recognized as a word, and then types
correct word of "remember" to replace "rene" with "remember."
However, unlike in the embodiments of FIGS. 3 and 4, the user does
not input a space after "rene," and continues to input correct
characters in order to to revise type error. Considering the
originally intended word, "remember" and actually input characters
"rene," the first and second characters do not involve errors, and
the error is made at the third character, thus the user inputs from
the third character of the intended word, "remember," i.e., to the
last character of the intended word. Accordingly, the user's input
is now "I renemember." At the state that the input of "renemember"
is complete as a word, it is determined whether such a word is in
the library. In case the word is not listed in the library, it is
determined whether there exist character substrings within the
word, which are highly correlated with each other. In the present
embodiment, character substrings "ne" and "me" in the word of
"renemember" are determined to be highly correlated with each
other, and then, it is determined whether a word, in which the
previously input character substring, "ne," is replaced by the
later input character substring, "me," exists in the library. Since
the result of such replacement is "remember," and this word is in
the library, the previously input character substring, "ne," is
replaced by the later input character substring, "me," to change
the input of "renemember" to "remember."
[0043] While the user inputs "renemember," the method of
determination of the correlation between character substrings is
described below. Here, among various methods for correlation
determination, focus is made on the high possibility of erroneous
input of a key transversely abutting the intended key in a keyboard
layout. That is, when "s" is input, the possibility that this input
has been erroneously made instead of intended input of "a," which
abuts "s" in the left, or instead of intended input of "d," which
abuts "s" in the right, is considered. When the first character "r"
is input, there does not exist another character to be compared
with it. When "re" is input, the first character "r" and the second
character "e" abut each other in a keyboard layout, thus the
possibility that "e" has been input in a revision intention is
considered, accordingly a revision candidate "A" is registered.
When "ren" is input, the third character "n" does not abut first
character "r" and second character "e," thus additionally candidate
is not registered. When "rene" is input, the fourth character "e"
abuts the first character "r," thus the possibility that the fourth
character "e" has been input in an intention to revise "ren" is
considered, a revision candidate "B" is registered. When "renem" is
input, the fifth character "m" and the third input "n" abut each
other, thus the possibility that the input of the fifth character
"m" has been made out of an intention for revising "ne" is
considered, accordingly a revision candidate "C" is registered.
Also, the possibility that the input of fifth character "m" has
been made out of an intention for revising "en" to "em" is
considered, accordingly, a revision candidate "D" is registered. At
this point, for the revision candidate "B," the input of fourth
character "e" should be followed by "e" in order to decide that the
fourth input "e" has been made out of an intention for revising
"ren," but the fourth character "e" is actually followed by "m,"
thus the revision candidate "B" is dropped from the revision
candidates. When "reneme" is input, the first character "r" and the
sixth character "e" abuts each other, thus the possibility that the
input of the sixth character "e" has been made out of an intention
for revising "renem" is considered, accordingly a revision
candidate "E" is registered. At this point, for the revision
candidate "C," the possibility that the user intention revising
"ne" to "me" becomes higher, thus the revision candidate remains as
a revision candidate. Further, for the revision candidate "A," the
character substring of "eneme" should be in the library in order to
decide that the second input "e" has been made out of an intention
for revising the first input "r," but the character substring
"eneme" is not in the library, thus the revision candidate "A" is
dropped from the revision candidates. When "renemem" is input, the
seventh character "m" and the third character "n" abut each other,
thus the possibility that the input of the seventh character "m"
has been made out of an intention for revising "neme" is
considered, accordingly, a revision candidate "F" is registered.
Further, the possibility that the input of the seventh character
"m" has been made out of an intention for revising "en," which are
second and third characters, to "em," which are sixth and seventh
characters, is considered, accordingly, a revision candidate "G" is
registered. At this point, for the revision candidate "E," which
considers possiblity of revision of "renem" the system determines
that the input of sixth and seventh characters, "em," is not made
out of an intention for revising "renem," accordingly, the revision
candidate "E" is dropped from the revision candidates. When
"renememb" is input, the eighth character "b" and the third
character "n" abut each other, thus, the possibility that the input
of the eighth character "b" has been made out of an intention for
revising "nemem" is considered, accordingly, a revision candidate
"H" is registered. At this point, for the revision candidate "F,"
which considers possibility of revision of "neme", the system
determines that the input of seventh and eighth characters, "mb,"
is not made out of an intention for revising "neme," accordingly
the revision candidate "F" is dropped from the revision candidates.
Further, for the revision candidate "G," which considers
possibility of revising "en," which are second and third
characters, to "em," which are sixth and seventh characters, the
fourth character "e" and the eighth character "b" should be at
least in a mutually abutting relationship, but this condition is
not satisfied, accordingly, the revision candidate "G" is dropped
from the revision candidates. When "renemembe" is input, the ninth
character "e" and the first "a" abut each other, thus, the
possibility that the input of the ninth character "e" has been made
out of an intention for revising "renememb" is considered,
accordingly, a revision candidate "I" is registered. When
"renemember" is input, the tenth character "r" abuts the second,
fourth, sixth and ninth character "e," accordingly revision
candidates "J," "K," "L," "M" are registered. At this point, for
the revision candidate "H," which considers possibility of revision
of "nemem," the system determines that the input of "ber" is not
made out of an intention for revising "nemem," accordingly the
revision candidate "H" is dropped from the revision candidates.
Finally, when a space is input, for the revision candidate "I,"
which considers possibility of revision of "renememb," it is clear
that "renememb" cannot be revised to "er," accordingly the revision
candidate "I" is dropped from the revision candidates. For the same
reason, the revision candidates, "J," "K," "L," "M" are also
dropped from the revision candidates. Now, the remaining revision
candidates are "C," which considers possibility of revision of
third and fourth characters "ne" to fifth and sixth characters "me"
and "D," which considers possibility of revision of second and
third characters "en" to fourth and fifth characters "em." Since
the revision result according to revision candidate "C" is
"remember," which is identical to the revision result according to
revision candidate "D," and the revised word, "remember," is in the
library, the revision is finally completed. It should be noted that
the above-described method of determination of the correlation
between character substrings is exemplary, thus any modification
can be practiced in the implementation of the present
invention.
[0044] FIG. 6 depicts an embodiment in which user directly provides
a revision command according to the third means. The user intended
to input "I remember," but at the state that the actual input is
made as "I rene," the user recognizes that the first and second
characters were correctly input, but a type error occurred from the
third character. Accordingly, the user resumes inputting from the
third character without inputting a space after "I rene." Now, the
user's input is "I renemember." The revision command is made
thereafter. According to the revision command, which can be
practiced variously as described above, a correlation between
character substrings within the input word is calculated. In the
present embodiment, character substrings "ne" and "me" in the word
of "renemember" are determined to be highly correlated with each
other, and then, it is determined whether a word, in which the
previously input character substring, "ne," is replaced by the
later input character substring, "me," exists in the library. Since
the result of such replacement is "remember," and this word is in
the library, the previously input character substring, "ne," is
replaced by the later input character substring, "me," to change
the input of "renemember" to "remember." For the method of
correlation calculation among character substrings, the same method
as described in the above embodiment of FIG. 5 can be used.
[0045] First to third means according to the present invention have
been described in the above. Each means can be practiced
independently and it is also possible for each mean to be
implemented in combination with other means.
* * * * *